XFree86-4.3.0-44 build fails in xc/programs/bitmap

[sean darcy]

I'm trying to build XFree from development. It fails here:

gcc -m32 -o bitmap -march=athlon-xp -m3dnow -msse -mmmx -mfpmath=sse -Os  
-fno-strict-aliasing -pip e -ansi -pedantic -Wall -Wpointer-arith -Wundef    
  -L../../exports/lib   BitEdit.o CutPaste.o Grap hics.o ReqMach.o Bitmap.o  
                  Dialog.o Handlers.o -lXaw -lXmu -lXt -lSM -lICE -lXpm - 
lXext -lX11     -lm -Wl,-rpath-link,../../exports/lib
../../exports/lib/libXaw.so: undefined reference to `.L91'
collect2: ld returned 1 exit status
make[4]: *** [bitmap] Error 1
make[4]: Leaving directory 
`/usr/src/redhat/BUILD/XFree86-4.3.0/xc/programs/bitmap'

Is this the correct list for build issues?

thanks

sean

_________________________________________________________________
Share holiday photos without swamping your Inbox.  Get MSN Extra Storage 
now!  http://join.msn.com/?PAGE=features/es

[Alexander Dalloz]

Am Do, den 27.11.2003 schrieb sean darcy um 18:06:
> I'm trying to build XFree from development. It fails here:
>
> gcc -m32 -o bitmap -march=athlon-xp -m3dnow -msse -mmmx -mfpmath=sse -Os  
> -fno-strict-aliasing -pip e -ansi -pedantic -Wall -Wpointer-arith -Wundef    
>   -L../../exports/lib   BitEdit.o CutPaste.o Grap hics.o ReqMach.o Bitmap.o  
>                   Dialog.o Handlers.o -lXaw -lXmu -lXt -lSM -lICE -lXpm - 
> lXext -lX11     -lm -Wl,-rpath-link,../../exports/lib
> ../../exports/lib/libXaw.so: undefined reference to `.L91'
> collect2: ld returned 1 exit status
> make[4]: *** [bitmap] Error 1
> make[4]: Leaving directory 
> `/usr/src/redhat/BUILD/XFree86-4.3.0/xc/programs/bitmap'
>
> Is this the correct list for build issues?

If you build test packages I think so.
> thanks
>
> sean

Btw. to set "-march=athlon-xp -m3dnow -msse -mmmx -mfpmath=sse" as gcc
3.3 compiler options the 3dnow, sse, mmx and -mfemath=sse calls are
redundant as the -march setting to athlon-xp will choose that all.

Alexander

[Mike A. Harris]

On Thu, 27 Nov 2003, Alexander Dalloz wrote:
> > I'm trying to build XFree from development. It fails here:
> >
> > gcc -m32 -o bitmap -march=athlon-xp -m3dnow -msse -mmmx -mfpmath=sse -Os  
> > -fno-strict-aliasing -pip e -ansi -pedantic -Wall -Wpointer-arith -Wundef    
> >   -L../../exports/lib   BitEdit.o CutPaste.o Grap hics.o ReqMach.o Bitmap.o  
> >                   Dialog.o Handlers.o -lXaw -lXmu -lXt -lSM -lICE -lXpm - 
> > lXext -lX11     -lm -Wl,-rpath-link,../../exports/lib
> > ../../exports/lib/libXaw.so: undefined reference to `.L91'
> > collect2: ld returned 1 exit status
> > make[4]: *** [bitmap] Error 1
> > make[4]: Leaving directory 
> > `/usr/src/redhat/BUILD/XFree86-4.3.0/xc/programs/bitmap'
> >
> > Is this the correct list for build issues?
>
> If you build test packages I think so.
> > thanks
> >
> > sean
>
> Btw. to set "-march=athlon-xp -m3dnow -msse -mmmx -mfpmath=sse" as gcc
> 3.3 compiler options the 3dnow, sse, mmx and -mfemath=sse calls are
> redundant as the -march setting to athlon-xp will choose that all.

Athlon reorders instructions internally anyway, the net result
generally being that it makes no difference what CPU instruction
scheduling is chosen, as the chip does it's own thing internally.

I'd be kind of surprised if compiling with all those options 
yields any real world performance gains anyway.  Placebo effect 
perhaps...  ;o)

[Ron Kuris]

Mike A. Harris wrote:
> Athlon reorders instructions internally anyway, the net result
> generally being that it makes no difference what CPU instruction
> scheduling is chosen, as the chip does it's own thing internally.
>
> I'd be kind of surprised if compiling with all those options 
> yields any real world performance gains anyway.  Placebo effect 
> perhaps...  ;o)

The gentoo folks would definitely have you believe otherwise.  A friend 
of mine swears that rebuilding everything using the -march=athlon-xp 
results in a much more responsive X subsystem.

I would be very curious to see a Fedora installation completely rebuilt 
xp-only.

Ron

[Michael Schwendt]

On Fri, 28 Nov 2003 19:26:29 -0800, Ron Kuris wrote:
> Mike A. Harris wrote:
> > Athlon reorders instructions internally anyway, the net result
> > generally being that it makes no difference what CPU instruction
> > scheduling is chosen, as the chip does it's own thing internally.
> >
> > I'd be kind of surprised if compiling with all those options 
> > yields any real world performance gains anyway.  Placebo effect 
> > perhaps...  ;o)
>
> The gentoo folks would definitely have you believe otherwise.  A friend 
> of mine swears that rebuilding everything using the -march=athlon-xp 
> results in a much more responsive X subsystem.

The funny thing about the "Gentoo folks" is, their perceived performance
improvement is not reproducible by all Gentoo folks. After wasting many
hours on recompiling stuff with modified compiler options and a largely
different set of build dependencies, not everyone finds the system to be
"much more responsive". Those, that do, should attempt at proving
performance increases with benchmark studies.

--

[Michal Zeravik]

Well, what should one do? Download XFree-* srpms, rebuild and reinstall?

michalz

Ron Kuris wrote:
> Mike A. Harris wrote:
> > Athlon reorders instructions internally anyway, the net result
> > generally being that it makes no difference what CPU instruction
> > scheduling is chosen, as the chip does it's own thing internally.
> >
> > I'd be kind of surprised if compiling with all those options yields 
> > any real world performance gains anyway.  Placebo effect perhaps...  ;o)
>
> The gentoo folks would definitely have you believe otherwise.  A 
> friend of mine swears that rebuilding everything using the 
> -march=athlon-xp results in a much more responsive X subsystem.
>
> I would be very curious to see a Fedora installation completely 
> rebuilt xp-only.
>
> Ron
>
> -- 
> fedora-test-list mailing list
> [e-mail]
> http://www.redhat.com/mailman/listinfo/fedora-test-list

[Jakub Jelinek]

On Thu, Nov 27, 2003 at 12:06:26PM -0500, sean darcy wrote:
> I'm trying to build XFree from development. It fails here:
>
> gcc -m32 -o bitmap -march=athlon-xp -m3dnow -msse -mmmx -mfpmath=sse -Os  
> -fno-strict-aliasing -pip e -ansi -pedantic -Wall -Wpointer-arith -Wundef   
>  -L../../exports/lib   BitEdit.o CutPaste.o Grap hics.o ReqMach.o Bitmap.o  
>                  Dialog.o Handlers.o -lXaw -lXmu -lXt -lSM -lICE -lXpm - 
> lXext -lX11     -lm -Wl,-rpath-link,../../exports/lib
> ../../exports/lib/libXaw.so: undefined reference to `.L91'
> collect2: ld returned 1 exit status
> make[4]: *** [bitmap] Error 1
> make[4]: Leaving directory 
> `/usr/src/redhat/BUILD/XFree86-4.3.0/xc/programs/bitmap'
>
> Is this the correct list for build issues?

No.  If you suspect a bug in the compiler (which this
looks like it), find out which libXaw.so object has undefined .L91
symbol and file a bugreport in http://bugzilla.redhat.com/bugzilla/
with the details (gcc command line used to compile that object
with .L91 undefined reference (use e.g. nm -u to find out), preprocessed
source (add -save-temps to gcc options, rerun it and attach the resulting
.i file) and exact gcc version you used).

Jakub